lean manufacturing

Lean Manufacturing & Kanban

CLOSE

Introduction

Overview

Much of the Japanese success in international markets has been the result of the adoption of lean manufacturing, a concept first introduced by the Toyota Motor Corporation. Lean manufacturing is a total system encompassing product design, equipment selection, materials management, quality assurance, line layout, job design, and productivity improvement. The Toyota Production System led to higher quality, lower cost, and substantially less labor time per vehicle than was achieved by Toyota's international competitors.

The production/material control system associated with lean is a "kanban" system. This course focuses on the broader philosophical aspects of lean manufacturing and describes kanban systems in detail. It also examines implementation issues as well as the strengths and weaknesses of kanbans. Finally, it will introduce alternatives to kanban inventory management, including CONWIP and Drum-Buffer-Rope, as part of a Theory of Constraints (TOC) approach to manufacturing management. To this end, principles of managing bottlenecks will be illustrated.

This course is designed for anyone who deals with production planning and control, particularly in high volume manufacturing or where there is a significant bottleneck.

Ohno, Taiichi 1988. Toyota Production System: Beyond Large-Scale Production. 1st ed. New York: Productivity Press, Inc.

Course Objectives

By the end of this training course, you will be able to:

·         Describe the six principles of lean manufacturing

·         Name the prerequisites of kanban or pull system implementation

·         Define cycle time and its five elements

·         Describe how pull systems differ from traditional push or schedule based manufacturing systems in operation and performance

·         Describe a “trial and error “ approach to sizing kanbans

·         State the equation relating cycle time, WIP and ship rate

·         Calculate kanban sizes given the kanban formula and appropriate information

·         State the five steps of the pull line design process

·         Balance the work load of a simple assembly process given a specific number of workstations and relevant task information

·         Describe how similarity of work content allows mixed model production on kanban lines

·         Define CONWIP and describe how it is different from a kanban system

·         Identify several situations or conditions where CONWIP can be effectively applied or avoided

·         Define a bottleneck as it applies to production operations

·         Define transfer batching and describe how it reduces lead time and bottleneck idle tim

A Note on “Lean” Terminology

Few fields are as prolific in their generation of terms with similar meanings as manufacturing. The following general conventions will be adhered to in this training course:

·         Lean or lean manufacturing. This is the general philosophical approach to operations management embodied in the six principles of lean.

·         Toyota Production System. This is a specific set of practices implemented by Toyota and documented by Tauchi Ono.

·         Just-in-time (JIT) production. This term represents an ill-defined set of operating methods including set up reduction, lot size reduction, kanban systems, and rate based scheduling. As a descriptive term, it is not very specific. Although it is used extensively in some organizations, it has attracted a degree of negativity due to early failed implementations.

·         Pull production. This is also a term for the same set of operating methods included in JIT. However, this term has not garnered as negative a connotation and is more descriptive of the intent to tie production to customer demand.

·         Kanban system. This term refers specifically to the mechanisms and operating rules that manage inventory movement between workstations on a pull production line. A kanban is the Japanese word for sign or signal. A kanban is a signal for the previous operation to provide a specific quantity of additional material. A kanban can be a physical container designated to hold a certain quantity.

In industry, these terms are all used somewhat interchangeably. In this training course we will attempt to adhere to the conventions described above.

Womack, J. P. and Jones, D. T. 1996. Lean Thinking: Banish Waste and Create Wealth in Your Corporation. 1st ed. New York: Simon & Schuster. Ohno, Taiichi 1988. Toyota Production System: Beyond Large-Scale Production. 1st ed. New York: Productivity Press, Inc.

PRINCIPLES OF LEAN MANUFACTURING

Overview

The basic philosophical principles of lean manufacturing were effectively captured by Womack and Jones in their book “Lean Thinking,” (1996). They provide a context for a more in depth discussion of using a kanban system to control the flow of inventory through a production system as one way to operationalize lean principles. Note that although a key focus of lean is “doing with less” as might be inferred from the term lean, these principles focus on value delivery. Value creation is as important as cost reduction.

Below are the principles of lean manufacturing:

Precisely Specify Value by Specific Product
Value is specified in terms of what the customer wants or needs. It is often very different from the way cost is measured in accounting.

Identify the Value Stream for each Product
The production process is viewed as a series of steps where value is sequentially added to the produc

Let the Customer Pull Value from the Producer
Product and work in process is only generated to satisfy a specific demand from the customer. The demand creates a “pull” for product that is transmitted sequentially through each workstation from the end of the production line to the beginning. Product is built to a pull signal, not “to schedule”.

Make Value Flow without Interruptions
Anytime a product waits or is delayed in the production process is a waste of time and resources.

Pursue Perfection
A lean manufacturing system strives for continuous improvement. This is both a competitive and operational imperative. Lean systems do not run well in highly variable and mishap filled environments.

Although not stated by Womack and Jones, most experts agree that the following is also a principle of lean manufacturing:

Empower Teams
The cooperation of workers and management is necessary for a lean system to run effectively. Everyone needs to feel ownership for the proper functioning of the production line and, ultimately, satisfying the customer. Lean systems leverage the axiom that “people doing the work know best how to improve it”.

Seven Wastes – “Muda”

One way lean systems address the concept of identifying value is by finding those things that definitely do NOT add value. Taiichi Ono, author and key developer of the "The Toyota Production System"(1988), provides a list of seven forms of waste, or “muda” as it is known in Japan. Select each item to view an example.

1.     Defects (in products)
Example – Anything that does not meet specification.
 

2.     Overproduction of goods not needed
Example – Warehouses of material waiting for customer orders
 

3.     Inventories of goods waiting further processing or consumption
Example – Pallets material sitting around a workstation when units can only be processed one at a time
 

4.     Unnecessary processing
Example – Counting inventory at each workstation
 

5.     Unnecessary movement (of people)
Example – Retrieving tools from a central tool crib that are regularly used at a workstation
 

6.     Unnecessary transport (of goods)
Example - All lathe machines are grouped together rather than located based on product/process flow
 

7.     Waiting (by employees for process equipment to finish an upstream activity)
Example – Unbalanced workloads between workstations cause some workers to be very busy while others idle
 

Inventory as Buffer –“Lowering the Water Level”

The dynamic interaction between lean manufacturing systems and inventory is often portrayed with a “lowering the water level” analogy. Inventory acts as a buffer in manufacturing against problems and inefficient operations. In the analogy, water represents inventory and the ship riding on the water symbolizes the production line. The water covers rocks or problems. As the water, or inventory level is lowered, the rocks are exposed. For lean to operate smoothly the rocks must be addressed.

The Lean Tool Box - Prerequisites of Pull Production

In order to avoid problems as inventory levels come down, several activities or tools are important prerequisites of pull production:

·         An awareness of waste and commitment to the principles of continuous improvement. This awareness and commitment must be present at all levels of management from operators to the highest management in the company.

·         Team based improvement activities. Teams must be established and be capable of applying continuous improvement tools such as brainstorming and cause and effect diagramming.

·         Process measures. Performance should be measured and rewarded based on process improvement and the elimination of waste. Traditional financial measures can conflict with lean principles.

·         Factory housekeeping – visual order, displays, and signals. The Japanese developed an approach to factory housekeeping called “5S”. It is a process for systematically organizing and maintaining the workplace so unnecessary items are eliminated, tool and material locations are standardized, and everything is kept spotlessly clean.

·         Quick changeover activities. Pull production necessitates small lot or batch sizes. If the setup time for each batch is long, then pull production will be very inefficient.

·         A multi-task trained workforce. Because inventory is only built when and if needed, operators must “flex” to where the work is. Operators that are only trained in one process step become very costly if production at that step is not required.

·         Cell design. Proximity greatly simplifies pull production. The closer sequential process steps are, the easier it is to communicate production demands. It is also easier to understand customer (next operation) quality requirements. Production operations with traditional functional departments should be re-laid out to enhance process flow.

Productivity Press Development Team, 2002. Pull Production for the Shopfloor.1st ed. New York: Productivity Press

Typical Improvements with Lean

The implementation of lean manufacturing systems can lead to dramatic improvements in operations performance. The following chart shows typical ranges of improvement achieved by operations that focus on lean manufacturing.

Thomas, Philip R. 1990. Competitiveness Through Total Cycle Time: An Overview for CEO's. 1st ed. New York: McGraw-Hill

Topic Summary

In this topic, the principles and concepts of lean manufacturing were introduced to provide the context and motivation for implementation of kanban or pull production systems. Lean manufacturing is both a philosophy and set of operating principles that focus on making the value delivery process as free of waste as possible.

The seven wastes targeted by lean implementers were enumerated. Lean systems operate with very little inventory to buffer against production problems. Therefore, certain capabilities should be considered as prerequisites to successful implantation of a pull production system, namely commitment to continuous improvement, teamwork, process focused measures, visual management, quick changeovers, cross-training, and cellular layout.

CYCLE TIME

Overview

One of the main performance measures lean manufacturing, and kanban systems in particular, act upon is TIME. In this topic, we will examine production cycle time as a critical, and much overlooked, measure of production line performance. Production cycle time is the total time it takes for one unit of production to complete the production process. This includes run time, any queue or delay time a unit might encounter, any assurance processes that may be required, and any move time.

Production cycle time is the total time taken to perform the entire set of operations necessary to complete one unit of production.

The “Any Process Cycle Time Model” shows the five elements of production cycle time. It is called the “any process” model because any process can be broken down into a serial combination of these five elements.

Five Elements of Cycle Time

The five elements of cycle time are described below:

·         Run time. Time spent adding value to the product. Work is being performed. Value is determined in the eyes of the final customer and not in terms of internal accounting conventions. Not all work adds value. In most traditional production processes less than 10% of the cycle time is run time.

·         Queue time. Time the material spends waiting to be worked on. The material has already been delivered to the next process operation, but work has not started. Queue time can be due to units having to wait their turn behind other work-in-process (WIP) or waiting for a setup to be completed. This is usually the biggest component of cycle time.

·         Delay time. Time the unit waits at the workstation after processing when it could be moved to the next workstation. Delays can arise from the human tendency to wait for a reasonable batch size before transporting units to the next work center. Reducing the distance between workstations or increasing the frequency of transport reduces the average delay time.

·         Assurance time. Time spent working on the product, but NOT adding value. Inspections, cleaning, counting, and status reporting are examples of non-value adding, assurance activities. It is important to point out that “have to” does not equal value added. For example the parts “have to” be cleaned, but this is because the process is long and dirty and the parts are exposed to contamination. The process is not capable without the cleaning step. A continuous improvement effort could potentially eliminate the need for cleaning. The same is true of inspections. However, some non-value adding activities cannot be readily removed. Think of the “assurance” designation as making the step a candidate for elimination.

·         Move time. Time spent transporting the material from one operation to the next.

The 24/7 Clock Standard

Although some operations managers will argue violently against this, cycle time is best measured against a 24 hour, 7-day clock. Customers pay for a product. They do not care that an operation only works one shift or workers need a couple of days off every week. Inventory cost does not care either. Working capital tied up over the weekend is still incurring finance charges.

Of course, operations need to make practical decisions about their work schedules. The point of using a 24/7 clock is so management can make rational tradeoffs in making decisions on how to prioritize their cycle time reduction activities.

Cycle Time: “A Cash Machine for Companies”

Why should mangers get excited about cycle time reduction as a focus for continuous improvement? In the 70’s and 80’s, cost and quality were the competitive priorities highest on everyone’s list. Cycle time has some unique and important relationships to other manufacturing fundamental measures that the manufacturing community is only now beginning to understand and appreciate.

Ed Heard, a consultant who worked extensively with Motorola during the early development years of their 6-sigma quality and cycle time reduction culture, proposes a relationship diagram that illustrates why cycle time is the “fundamental’s fundamental”:

It is relatively easy for any manager or accountant to argue for the right half of the diagram:

·         Lower operating costs → more cash

·         More satisfied (happy) customers → more sales revenue → more cash

·         Less working capital required → more cash

The next several screens, however, will investigate the impact of cycle time on the measures on the left-hand side of the diagram.

Cycle Time Reduction’s Affect on Other Measures

Cycle time has the following impacts.

Non-value added activities

The principle mechanism to remember here is that during the entire time a unit is in process (cycle time) it is at risk. It is at risk for loss, handling damage, contamination, and obsolescence. To combat these risks, the operation must do work in the form of inventory tracking and counting, inspections, storage in protective containers, etc. This “assurance” work is non-value added. As the cycle time goes down, the need for assurance of work goes down. For example, inventory with a cycle time of 10 days needs tracking at every process step. If the cycle time is only 5 hours, why not track only enter and exit steps?

Quality

The principle mechanism here is the reduction in the time between cause and effect. Put more simply, it is the reduction in the time between when something bad happens and when the problem is discovered. Because cause and effect are close together, it is more likely the causal relationship will be fresh and apparent. Think how hard it is to remember what went wrong on a seemingly routine shift even 2 days ago. By reducing the cycle time, the ability to solve quality problems improves and, therefore, quality itself tends to improve.

Delivery

Less time between start and finish means we can execute delivery faster. However, in manufacturing, delivering on-time is as important as delivering fast. Reducing cycle time and choosing to keep delivery lead-times constant establishes a “time buffer” between the time an order is completed and the time it needs to be delivered. Typically, if cycle time reduction is in terms of weeks, the benefit is given to marketing and the standard quoted lead-time is reduced. Manufacturing also delays its start by the same amount. Smart managers consider retaining at least a portion of the time buffer to improve their on-time delivery performance. In ether case, reducing cycle time either shortens delivery lead-time, improves on-time performance, or both.

Inventory

We will talk in more detail about the explicit relationship between time and inventory later in this course. For now, we can look at the relationship between inventory and cycle time with a pipeline analogy. Inventory forms a pipeline from the start of a process to its finish. The longer the pipeline (cycle time) the more inventory is required to fill the distance from start to finish. If the cycle time is reduced, we have a shorter pipe and less inventory.

Other Measure’s Affect on Cycle Time

Note that the relationship diagram also has arrows pointing from non-value added, quality, delivery, and inventory to cycle time.

This was not in Ed Heard’s original diagram, but it is the key to cycle time's robustness as a focus for continuous improvement in business. The effect of these bi-directional arrows is to create a reciprocal relationship where improvement in one drives an improvement in the other. This reciprocal relationship is what makes cycle time so compatible and powerful as an improvement focus.

Other Measure’s Affect on Cycle Time - Continued

We have already looked at how cycle time reduction drives improvement. Let us now see how improvement in non-value added, quality, delivery, and inventory improves cycle time:

·         Non-value added activities. Reducing the number and duration of non-value added activities reduces the amount of work time to complete the process → less cycle time.

·         Quality. Quality problems cause work in the form of inspections. They also cause delays. If the quality problem is bad enough that scrap results, then more work must be done to replace the units. The time invested in the original units is not recoverable. Reducing quality problems → less cycle time.

·         Delivery. The role that delivery plays in cycle time reduction is interesting. It can be argued that if product is consistently delivered on time and at quantity, customers are less likely to place back-up, or ‘phantom’, orders they may cancel if the original order comes through as expected. These phantom orders induce more queue time and steal valuable capacity from the product. Providing dependable delivery → less cycle time.

·         Inventory. Reducing inventory is the result we target when we implement inventory control systems such as kanban. As we will see in the next topic, inventory is directly proportional to queue or waiting time. By forcing inventory down the result is → less cycle time.

Topic Summary

This topic presented the importance and benefits of reducing production cycle time, defined as the total time it takes to produce a unit of product from start to finish. It includes the five cycle time elements: run time, queue time, delay time, assurance time, and move time. A 24 hour, 7-day clock is used to measure cycle time because customers do not pay for idle time and working capital tied up does not respect weekends.

As was shown, cycle time is a key measure that affects overall business performance. Reducing cycle time eliminates non-value adding activities, improves quality and delivery, and reduces inventory. In turn, improving these measures also reduces cycle time. It is this reciprocal relationship between cycle time and these key manufacturing measures that makes cycle time such a robust focus for continuous improvement.

KANBAN SYSTEM OPERATIONS

Overview

The previous two topics in this training course have discussed how lean manufacturing is a philosophy of focusing on value creation and the elimination of waste in the manufacturing process and how cycle time is a key measure for the level of improvement. This topic introduces one specific mechanism for achieving and maintaining low production cycle times. The mechanism is called “kanban”. Kanban is a Japanese tern that means “sign or signal”.

In manufacturing, it is used to describe a production system whereby the customer demand, or need, is signaled back through successive operations to the beginning of the production process. In this manner only material that is needed to satisfy a particular customer order is allowed to enter the production process. Kanban is the primary mechanism of a pull production process. “Only proceed if there is a need” is the basic credo of a kanban process.

Traditional push and kanban systems will be demonstrated in this topic. Additionally, two approaches to kanban system design will be described.

Traditional “Push” Manufacturing

Traditional, centralized, schedule-based, or “push” manufacturing provides each work center in a process with an output target. Operators endeavor to meet or exceed those targets. The focus is on output, many times to the detriment of quality, teamwork, and inventory targets.

A common characteristic of these systems is that they tend to overproduce unneeded units and under produce those that are needed.

“Pull” Manufacturing

Kanban systems are sometimes called “pull” manufacturing because customer demand creates a need or pull transmitted back through the production line. The need is determined locally and is not dependent on a central planning system.

A kanban signal can take many forms. It can be as simple as an empty square on a workbench with the governing rule, “fill with one and only one unit when empty”. By removing the unit from the square the receiving work center creates the kanban signal – EMPTY SQUARE. FILL ME! Other typical forms of kanban are cards or computer displays. In whatever form, kanbans have the following characteristics:

1.     Demand is locally determined by the next workstation.

2.     A fixed quantity or “queue limit” is associated with each kanban.

3.     A specific product, assembly, sub-assembly, or component is required

Topic Summary

The past topic showed the difference between traditional, schedule-based push systems and a pull system that allows the customers’ demand to be transmitted back through the production line from the end to the beginning. While output is generally the same in both systems, the pull system achieves the same throughput with shorter cycle time and fewer inventories.

Kanban is a very simple, systematic, and visual system. A kanban signal such as an empty square or card lets an adjacent upstream station know when more work is needed. An empty kanban can be interpreted as a workstation saying, “I am empty and need a fixed quantity of a particular type of material and no more!”

CALCULATING NUMBER OF KANBANS

Overview

In this topic, we will describe a formal process from designing a simple kanban system for a production line. So far, we have been using process steps with equal processing times and kanban sizes for ease of demonstration. Kanban sizes can actually vary due to differences in processing time between workstations, process characteristics, and material constraints.

The following topics will be addressed:

·         “Trial and error “ approach to sizing kanbans

·         Equations for cycle time, WIP and ship rate

·         How to calculate kanban sizes given the formula and appropriate information

Trial and Error Approach

Although this topic does include formulas and calculations, good old “trial and error” in kanban sizing is not out of the question. Kanban sizing is a good candidate for a trial and error approach because it is:

·         Robust with respect to the number of kanbans. With an adequate safety factor, having an extra kanban or two does not severely limit performance. Actually, if you start from the premise that most traditional manufacturing systems have too much inventory, just setting the initial kanban level at a significant fraction (e.g. one half might be a good place to start) of the exiting average inventory level is a practical place to start. Once capability is demonstrated at a particular level, the number of kanbans can be reduced as part of a continuous improvement process.

·         Easy to change. Kanban sizes are relatively easy to change. Often, there are no hard tooling or layout repercussions from changing kanban sizes. Kanbans may be changed from 5 to 4 to 3 and back to 4 as quickly as the new rules can be communicated. They are well suited for true on (production) line simulation experiments. This is not to say that kanban levels should be changed frequently. As with most manufacturing systems, kanban systems respond well to stability and practice.

Cycle Time and WIP Relationship

Cycle time and work in process (WIP) have a very simple and direct relationship. “It is calculated as the number of units of work in process inventory divided by the number of units processed in a specific period” (Thomas et al., 1990). In equation form:

Where:
C = Average Cycle Time for Work in process
W = Units of Work in process inventory
S = Units Shipped per time period

For example: if there are 5 units in process (W) and you process 1 unit per day (S), then the cycle time (C) equals 5 days. The next figure shows this pictorially. The check marks represent units of WIP:

It takes exactly 5 days for all the units to clear the system assuming all the process steps are capable of processing at least 1 unit a day.

Alternatively, if the ship rate is given in units per month and there are 5 units in process for a ship rate of 30 units/month than C = 5/30 months or 5 days for a 30 day month.

Cycle Time and WIP Relationship - continued

The relationship between cycle time and WIP is very powerful. Its strength lies in the ability to determine an average cycle time by knowing only the WIP and the total shipment for a period. Some limitations and assumptions must be understood in exploiting this relationship.

DETERMINATION IS INDEPENDENT
OF WIP LOCATION WITHIN SYSTEM

1.     Average cycle time. The first is the classic limitation of all averages. The equation results in an average cycle time. It says nothing about the max, min, or variation in cycle time for the time period examined. For example, the actual cycle time for the 30 units shipped could have been made up of 15 units that took 3 days and 15 units that took 7 days. We are only accessing the end result in that we know we shipped 30 units using 5 units of WIP.

2.     Average WIP. This leads to the WIP assumption. The WIP is also an average value. Over a narrow enough time period, the WIP level can be assumed constant. Therefore, at any given inventory snap shot, the WIP level recorded is close enough to the average value for that time period to calculate a representative cycle time value. Most inventory tracking systems can provide a true average WIP value.

3.     Location of WIP. It is also important to note that where the inventory resides between the start and stop points of the process is not significant. The formula treats the analysis as a closed system. Temporary bottlenecks within the systems are accounted for in the average. Production holds (management directed stopping of units) and violations of FIFO rules are captured to the degree that they affect the average cycle time.

Kanban Formula Introduction

We can now leverage the relationship between cycle time and WIP to develop a formula to calculate how many kanbans belong in front of each workstation. Kanbans should account for the following four factors:

·         Different replenishment times for each workstation (r). Not all workstations have the same amount of time between when an order (kanban signal) is received and delivered. Kanban formulas should account for replenishment time (l) of the supplying workstation.

·         Maximum operating demand rate of the line (d). The kanban needs to be large enough not to run out of material when the line is running at its maximum output rate (d). This is not necessarily the given work station, but it is determined by the workstation with the minimum output capacity (bottleneck).

·         Lead-time variation safety factor (s). A safety factor (s) needs to be added to the number of kanban to account for variation in lead-time from the supplying workstation.

·         Container size (c). Units delivered to a workstation may come in a standard or custom container with space for a fixed number of units. If the units are large, or do not have a container, a value of one can be used for the container size. It is also possible that a container can be specially designed to facilitate handling or interface with the line machinery.

Kanban Formula Explanation

Kanbans are simply queues of inventory that have a fixed, pre-designed, maximum size. They act as a cap on inventory trying to enter a process step -- material can only go into a kanban if it is not full. In addition, a kanban ensures the downstream process will have material when needed. In order to size a kanban so inventory between adjacent producing and consuming process steps is neither too much nor too little, the following relationship is applied:

Where:         k = number of kanbans or containers
                   d = demand rate
                   s = safety factor
                   c = container size
                   r = total replenishment time, which includes:

                p = process time to make a unit to fill the kanban container
                m = move time for producing to consuming workstation
                a = Set-up or assurance time for the producing workstation

Example – Calculating Kanban for Workstation B

Workstation A produces a product consumed at Workstation B. Workstation A takes 2 hours to set-up. It has a capacity of 200 units per hour. The customer requires a maximum rate of 110 units per hour. Workstation B has capacity to make 150 units per hour. It takes practically no time to move the parts from Workstation A to Workstation B since they are adjacent. The process design engineer considers a safety factor of .15 adequate to protect process B from variation in the replenishment time of Workstation A. The line operates three shifts per day, 7 days per week.

d (demand rate)                             = 110 units/hr
p (process time per container)        = 0.5 hrs (1/200 uph x 100 units/container)
m (move time)                               = 0
a (set-up time)                              = 2.0
r (replenishment time, total)           = 2.5 hours
S (safety factor)                            = 0.15
C (container size)                         = 100

The standard convention in calculating kanbans is to round up. Kanbans are counted in whole units. All empty kanbans signal the same number of units. Remember we do not want to run out of material, as might be the case if we rounded down or to the nearest whole number.

Example – Calculating Kanban for Workstation B - Continued

The next sequence of diagrams shows how the kanban system works for this workstation. Click the arrow to see how the kanban system will work under different situations.

Topic Summary

This topic introduced a very simple relationship between cycle time, WIP, and ship rate:

C = W/S

This relationship forms the basis for calculating kanban sizes. Kanbans are sized so that there is never too little or too much inventory at any one workstation along the production line. While there is a formula for the number of kanbans that uses demand rate, the lead-time of the supplying operation, and the kanban container size, trial and error is also a practical approach to determining how big a kanban should be.

PULL PRODUCTION LINE DESIGN

Overview

In the previous topic, we examined how to calculate a kanban quantity for an individual workstation, so that the process would neither have too much or too little inventory. In this topic, we will outline a basic design process to establish a kanban system from shipping to the acquisition of raw materials.

Information covered will include:

·         Five steps of the pull line design process

·         Methods of balancing the work load of a simple assembly process given a specific number of workstations and relevant task information

·         Ways that similarity of work content allows mixed model production on kanban lines

Pull Line Design Process

The basic line design steps are as follows. Click on each link for more details on the specific stages of the pull line design process.

1.     Flow chart the manufacturing process
Create a flow chart of all the work required to complete a product from the start of the production process to a completed unit.

2.     Calculate the design demand rate for the production line (D)
This is the output rate the line must be capable of producing to meet customer demand. It is calculated by dividing the targeted monthly volume by the number of workdays per month.

Example: Manufacturing and marketing agree that the production line needs to be capable of meeting a continuous maximum customer demand of 400 units a month. The plant works 20 days a month on average.

The design demand rate is calculated as:

3.     Calculate the operational “takt” time (t)
This is the targeted rate at which product must exit the line in order to meet the design demand rate. The work content of each workstation in the process is grouped so as not to exceed the operational takt time. The takt time is calculated by dividing the number of work hours in a day by the design demand rate:

Example: There are two 8-hour shifts per day, but allowing for breaks, training, and downtime the effective number of hours per shift averages only 7.5 hours. The design demand rate is 20 units/day.

To calculate operational takt time:


The reciprocal of the takt time is the operating demand rate (d) used in the kanban formula for each work station.

4.     Balance the assembly line
Group the process steps into workstations so that the precedence relationships in the flow chart are not violated and the work content is as close to the takt time as possible without exceeding it.

5.     Size the kanbans for each workstation
Apply the kanban formula to each workstation to calculate the required kanbans.

Results of Design Process

The result of this design process will be a line with the following characteristics:

·         The maximum inventory level will be fixed at the sum of the entire inventory in the kanbans. The inventory may be lower, but it will never be higher. This means inventory risk and cost are relatively fixed.

·         The total average cycle time will be relatively low and predictable. This allows manufacturing to make reliable commitments on deliveries.

·         The line will be capable of operating at any demand rate up to the design demand rate. Kanban systems have great downward flexibility, a drop in demand means a drop in pull. The line responds to changes in demand rate while holding inventory relatively constant.

Mixed Model Production

So far, we have been focusing on the single model case. Kanban systems can also be effective in the mixed model case under certain conditions. Mixed model manufacturing occurs when several models of a particular product or product family are built on the same production line. For example, a car manufacturer has the capability of building 2 and 4-door, economy and luxury models on the same production line under the same controlling set of kanban rules.

In order to use a kanban system under mixed model or product conditions, several conditions are generally required:

·         Products must have similar process flows, and must share the same routing through the workstations

·         Work content must be similar in that it requires similar tooling and operator skill sets

·         Work content must be executable within the takt time of the production line

·         Changeover time between models must be short, and any changeover or setup time must be accomplished within the workstation’s takt time

Mixed model production actually allows what on the surface seem to be quite different products to be built on the same production line. The underlying feasibility is determined by the similarity of the “work content”, not the products or their function. For example a company might build several varieties of electronic assemblies on the same line because they all go through a similar, “screw down, insert components, solder, clean, test” sequence.

Topic Summary

This topic introduced the five steps of the pull production line design process:

1.     Flow chart the process

2.     Calculate the design demand rate for the production line based on customer demand

3.     Calculate the operational “takt” time

4.     Balance the assembly line

5.     Size the kanbans for each workstation

The result is a line with fixed inventory risk and cost, low average cycle time, and the capability of operating at any level below the design demand rate. It is possible to use a kanban system under mixed model conditions as long as the models meet the line design criteria. The products should have similar product flows, work content, and takt times. In addition, the changeover time between models must be short enough to be accomplished within the takt time at any given workstation.

ALTERNATIVES TO KANBANS

Overview

Kanban systems work by keeping inventory constant and operating below a design demand rate. The inventory is evenly spread across the production line. Kanban systems work best when the workload and capacity of the individual workstations is relatively balanced. Although they can handle some variation in processing time, they work best in conditions of low variability and constant demand.

Two other alternatives to kanban will be explored in this topic, namely constant work in process or CONWIP and drum-buffer-rope from the Theory of Constraints approach to production control. As we will see, all three methods share similarities while offering unique features that make them attractive for particular situations.

Constant Work in Progress (CONWIP)

Constant work in process, or CONWIP, actually takes us back to where we started in leveraging the relationship between cycle time, WIP, and ship rate. If we are mainly concerned about maintaining a given maximum, overall inventory level and a predictable average cycle time, then we can devise a production rule that says, “Only start another unit into this sequence of process steps if one is completed.” A CONWIP system maintains a constant inventory, while allowing more flexibility of movement within a certain well-defined range of the process than a kanban system.

Arbitrary Distribution of WIP in a CONWIP system

In the diagram above, we can establish a CONWIP level of five by only starting a new unit if one ships out of the third step. Units can be at any one of the three steps, but there can be no more than five total. The cycle time is an average of 5 days and the inventory is constant at five units.

CONWIP systems typically use a limited number of special cards or containers. This creates a good physical control system – a new unit only starts if an empty container or card is available.

When and Where to Use CONWIP

CONWIP systems work well if one or more of the following conditions exist:

1.     Flow rate is low and does not warrant the overhead of implementing a kanban system.

2.     The process sequence is not standard, as may be the case in a job shop where the process sequence is dependent on a particular job.

3.     The process starts and stops often, as in engineering experiments where units are held for measurement or testing periodically.

4.     The process area is well defined or contained. This allows the number of WIP containers that enter or exit the area to be controlled. Walls and doors work well for this purpose.

5.     The process has rework or loop back processes where parts do not proceed predictably from start to finish. For example, a silicon wafer fabrication process may use a CONWIP system because the process loops back through a masking step multiple times before exiting the process, and the processing areas are not laid out in a sequential flow.

6.     The process is long and priorities change frequently on which units are most important.

7.     A specialized container or work holder exists that can be limited. For example, an electronic component assembly operation has special carriers called “magazines” that fit in the die attach and wire bonding machines.

Cautions with Respect to CONWIP

CONWIP is an excellent tool to use in maintaining inventory levels. It directly contradicts the common human tendency to start something new before completing what is already in process. Several cautions and negatives should be considered in the application of a CONWIP system:

·         Flexibility can also be interpreted as a lack of rules. CONWIP itself does not imply a priority system for lot selection. There is a tendency for operators to gravitate toward favorite operations or jobs if clear priorities are not established. This creates the need for a secondary priority system.

·         Progress is not always visible. One of the big advantages to a kanban system is that anyone can walk out on to the production line and see how close a particular work piece is to the end of the line. This is not always the case with CONWIP.

·         Cycle time variability can be very high. Although the average cycle time is consistent, the variation can be very high if the CONWIP system does not use a FIFO priority system. In a complex system with high CONWIP, some lots can be sidelined while others race through the process.

·         WIP can become unevenly distributed across the process. One or two stations can be buried in work while others are idle. This is a common occurrence when a machine breaks down and work piles up behind it. The good news is that CONWIP supplies a natural cut off mechanism. Once all the containers are in the process behind the breakdown, no more inventory can be started and put at risk.

Theory of Constraints (TOC) – Overview

The maximum throughput of both kanban and CONWIP systems is limited by the workstation or process step with the least capacity. This is commonly known as the “bottleneck”. In designing a kanban system we balance the line and try to make the capacity of each workstation as equal as practical. The line is run at an operational demand rate below the bottleneck capacity. In a CONWIP system the bottleneck exists somewhere in the process. It is usually easy to spot. It is the process step with most of the CONWIP containers piled up in front of it. In many industries, capacity management is the key to competitive advantage. The objective is to achieve the maximum output with the least resources.

A philosophy of manufacturing management that focuses on the management of bottleneck operations is known as Theory of Constraints (TOC). Bottlenecks are constraints on manufacturing throughput. TOC management was first introduced by Goldratt and Cox in their classic book The Goal (1984). The Goal put forward the proposition that the objective of any manufacturing operation is to make money. Money is being made if three conditions are being met:

·         Operating expenses are decreasing

·         Inventory is decreasing

·         Throughput is increasing

One of the key tenets of TOC is the identification and management of the bottleneck operation in a production process. Since the bottleneck determines throughput it is obvious that output will be good if the operational priority is to keep the bottleneck running. We will look at two mechanisms used in TOC to keep the bottleneck running while maintaining low costs and inventory. They are “drum-buffer-rope” and “transfer batching”.

Drum-buffer-rope

In order to describe the TOC drum-buffer-rope mechanism to manage a production line, we will use several concepts previously introduced.

·         The Drum. The first concept is that of takt time. Takt time is the “beat” of the production process. It is the time between subsequent units coming off the production line. In kanban line design, each workstation must be capable of work at or below the takt time. If the goal is to maximize the output of the system, why not use the “beat” of the bottleneck rather than the average time? The bottleneck becomes the “drum” that beats out the production time for the whole line. Every unit through the bottleneck signals a unit of output for the line and the need for another unit to be started.

·         The Buffer. The bottleneck or drum needs to run constantly. It certainly should never wait for material (WIP) to process. A “buffer” of WIP is maintained in front of the drum to make sure it never runs short of material to process. Theoretically, this is relatively easy to do because all of the process steps prior to the bottleneck have higher capacity. The trick is to make sure the inventory level does not grow too high. A CONWIP system in front of the bottleneck forms the buffer. It should be sized so that the queue in front of the drum is never zero.

·         The Rope. The ‘rope” is analogous to the boundary that is used in a CONWIP system. The rope is “tied” to a start operation several steps before the bottleneck. The rope is really a gate signal from the bottleneck, or drum, to the start operation to let another unit into the series of process steps before the bottleneck.

Drum-buffer-rope - Continued

The diagram to the right depicts a drum-buffer-rope system with Process Step 4 as the drum tied to Step 2 as the gating operation and a buffer inventory of 12 units:

A capacity diagram is a useful tool for identifying and communicating the relationship of the bottleneck’s capacity to the other process steps.

Transfer Batching

The drum-buffer-rope concept provides an overall framework for how to manage the line for both low inventory and high throughput. Transfer batching recognizes the fact the bottleneck operation should not have to wait for a unit’s completion, regardless of the original order or economic lot size of the upstream operations. A transfer batch is a subset of a large batch that is forwarded to the next operation before the original batch is completed. Transfer batching generally requires more administrative overhead than conventional constant batch manufacturing, but the timesavings on the bottleneck operation can be substantial.

Review the chart to the right, which shows a lot size of 100, with no transfer batch. Then, click next to see what happens when there is a transfer batch of 10.

Cautions with Respect to TOC

Theory of constraints and its associated control mechanisms, drum-buffer-rope and transfer batches, are similar in many ways to CONWIP and kanban systems. All strive to maintain inventories at lower levels than traditional schedule or push base manufacturing techniques. However, a few cautionary comments are warranted with respect to applying TOC concepts:

·         Bottlenecks can (and should) change. A phenomena called “roving bottlenecks” occurs when capacities are close to equal across workstations or process variability is high. Management by constraint then becomes highly problematic since the priority is always changing. The tendency to move towards equal capacities is actually aided by TOC’s drive to “break” the bottleneck. Once a bottleneck is broken, the new bottleneck has a capacity closer to the other workstations. A TOC approach works best in automated, capital-intensive operations that have one or two large, expensive pieces of equipment that retain bottleneck status.

·         TOC can give managers a false sense of control. TOC is a “tops-down” paradigm. Managers are encouraged to believe that by identifying a few key bottleneck operations they can control the whole process. In reality, many things can go wrong in even non-bottleneck operations that can keep an operation from achieving its goals.

Topic Summary

The CONWIP system maintains a constant inventory without setting restrictions on the individual workstations as in a kanban system. The CONWIP system is ideal in situations where the flow rate is low, the process is well contained, or where the process has steps that loop back.

The bottleneck is the workstation in the production line with the least capacity and determines throughput for the entire line. Knowing the bottleneck can be useful in that it acts as the “drum” in a drum-buffer-rope production control mechanism. The drum sets the pace for the production line, while the buffer is the WIP maintained in front of the bottleneck to ensure that it never runs short of material to process. The rope is really just a signal from the bottleneck to a gating operation that determines when to let another unit in to the buffer.

Transfer batching uses a batch that is a subset of a larger original batch and forwards it to the next operation before the original batch is completed. It recognizes a bottleneck operation should never have to wait for a whole batch’s completion. Using transfer batches also reduces the cycle time on the entire process.

COURSE SUMMARY

Summary

This training course examined new, alternative approaches to traditional schedule-based manufacturing control practices. Specifically four techniques were discussed in detail and shown to have advantages in different situations:

·         Kanban systems. Very effective in high volume repetitive processes that are labor and assembly intensive. Work can be balanced between workstations and a set of kanban work rules instituted so that WIP only moves in response to customer demand.

·         CONWIP systems. Maintain a fixed inventory level between specified process steps. CONWIP systems are most effective when process flows are low volume and unpredictable.

·         Drum-buffer-rope systems. Work best in situations where there is a well-defined, stable bottleneck. The bottleneck acts as a “drum” to meter material into the production line, thereby maintaining a fixed inventory level in front of the constraint operation. TOC based drum-buffer-rope systems are best employed in capital-intensive industries where throughput and capacity management are important competitive advantages.

·         Transfer batching. Method of ensuring that bottleneck operations are not idled. Small batch sizes are small and total cycle time is short.

All of the techniques in this training course share a preference for low, well-controlled inventory levels, small batch sizes and low changeover times. Every one of them can be effectively utilized in a well-run cost effective manufacturing operation.

Resources and Bibliography

The following resources were used to create this course, and are recommended for further information on the various topics covered.

Lean Manufacturing
Ono, T. 1988. Toyota production system: beyond large-scale production. Cambridge, Mass.: Productivity Press.
Womack, J. P., Jones, D. T., & Roos, D. 1991. The machine that changed the world: The story of lean production (How Japan's secret weapon in the global auto wars will revolutionize western industry) (1st Harper Perennial ed.). New York, NY: Harper Perennial.
Womack, J. P. & Jones, D. T. 1996. Lean thinking: banish waste and create wealth in your corporation. New York: Simon & Schuster.

Cycle Time Reduction
Hall, R. W. 1983. Zero inventories. Homewood, Ill.: Business One Irwin.
Heard, E. 1995. Cycle time - a cash machine for companies! Unpublished.
Thomas, P. R. & Martin, K. R. 1990. Competitiveness through total cycle time: an overview for CEOs. New York: McGraw-Hill.

Kanban System Design
Rother, M. & Shook, J. 1998. Learning to see (Version 1.2, June 1999 ed.). Brookline, MA: The Lean Enterprise Institute.
Rother, M., Harris, R., & Lean Enterprise Institute. 2001. Creating continuous flow: an action guide for managers, engineers and production associates (Version 1.0. ed.). Brookline, Mass.: Lean Enterprise Institute. 2002. Pull production for the shopfloor. New York: Productivity Press.

Theory of Constraints
Goldratt, E. M. & Cox, J. 1984. The goal: a process of ongoing improvement (2nd rev. ed.). New York: North River Press.
Goldratt, E. M. & Fox, R. E. 1986. The race. Croton-on-Hudson, NY: North River Press.
General Manufacturing Planning and Control Reference Vollmann, T. E., Berry, W. L., & Whybark, D. C. 1997. Manufacturing planning and control systems (4th ed.). New York: Irwin.

Content Author Information
Eric Olsen is a researcher, consultant, trainer, and doctoral student in Operations Management at Ohio State University. He has over 20 years of industry experience in engineering and manufacturing with companies like Caterpillar Tractor, Litton Industries, and Hewlett Packard. He has extensive experience implementing kanban, CONWIP, and TOC systems in pursuit of performance improvement. His current research interests are performance measurement systems and time-based competition. Contact information: 614 487-0382, olsen.54@osu.edu